InternationalJournald

Industrial
International Journal of Industrial Organization Organization
ELSEVIER 14 (1996) 181-201

Standardization, diversity and learning:

Strategies for the coevolution of technology and
industrial capacity
Paul A. David a'b'*, Geoffrey S. Rothwell a'*
aStanford University, Stanford, CA 94305-6072, USA
ball Souls College, Oxford, OX1 4AL, UK

Received 17 March 1993; accepted 10 October 1994

Abstract

This paper presents a model of industrial process standardization, diversity, and

learning. We show that when the effects of learning through diversity are strong, the
present value of long-run costs can be minimized with either complete standardiza-
tion or with complete experimentation where no two plants are similar in the early
stages of the industry. We also discuss the relevance of these and other analytical
results to the U.S. nuclear industry's standardization policies.

Keywords: Standards; Standardization; Diversity; Learning; Nuclear power

J E L classification: 0 3 2

* Corresponding author.
*This is a revised version of material originally developed by the authors in 'Performance-
based Measures of Nuclear Reactor Standardization', Center for Economic Policy Research
Publication 247 (6 June 1991), presented at the Third Meeting of the Conference Series on the
Dynamics of Large Technical Systems held in Sydney, Australia (1-7 July 1991). Our research
has benefited from the comments of Kenneth Arrow, Takeshi Amemiya, Raymond de Bondt,
W. Edward Steinmueller, Roland Sturm, and an anonymous referee. Financial support for this
work initially was provided as part of the project on 'Information and Organizational Impacts
on Productivity', under National Science Foundation grant IRI-8814179 to the High Technology
Impact Program, Center for Economic Policy Research, Stanford University. David also
acknowledges the subsequent support provided by the Economic and Social Research Council
(UK) Programme on Global Economic Institutions, under Grant L 12025 1003.

0167-7187/96/$15.00 © 1996 Elsevier Science B.V. All rights reserved

S S D I 0167-7187(95)00475-0
182 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

I. Introduction

Standardization has once again become a subject of lively concern in both

the business and the academic worlds. A half-century ago the cooperative
effort to reduce the variety of industrial product designs and to standardize
on what engineers thought to be 'the best' was able to draw some interest
even from those outside the narrow circle of technical experts and industrial
managers engaged in the 'rationalization movement' of the 1920s and 1930s.
Soon thereafter the topic slipped out of the consciousness, and off the
research agendas of economists and political scientists. The setting of
technical standards for products and production processes was held, by
those who remained aware of it at all, to be one of those arcane and tedious
matters best relegated for the attention of engineers, and, in the case of
safety standards, possibly also to lawyers expert in matters of product
liability and workmen's compensation law. During the past decade, how-
ever, standard-setting and standardization have made a great comeback.
Increasingly these are acknowledged to be issues the economic importance
of which is sufficiently great to warrant the concern of corporate executives,
research managers, public policy-makers in areas affecting technology and
industrial development, not to mention governmental regulators (see, for
example, David and Greenstein, 1990; OTA, 1992; and the extensive
literature cited therein).
What are the economic benefits from standardization, and what are its
costs? How do they balance over time? How much standardization is
optimal from the private viewpoints of producers and consumers? Is the
growth of markets and of profits in comparatively new lines of business
being retarded by the failure to establish timely standards? Are there
significant social welfare losses due to excessive production of variety and
failure to achieve standardization that would induce competitive entry, or
cost-reducing economies of scale? Or is it a good thing that standards are
not freezing the development of technology, by ossifying procurement
requirements or encouraging such intense competitive conditions as to
diminish the prospects that innovators would be able to recoup their
investments in research and development of still more powerful tech-
nologies? This is but a sampling of the questions that have surfaced in recent
discussions (see, for example, Kindleberger, 1983; Foray et al., 1991; and
David, 1994).
Interest in these and still other issues has been especially intense among
those concerned with the vending of equipment, the operation of networks,
and the provision of enhanced network services in the computer and
telecommunications industries. This hardly is surprising, in view of technical
compatibility standards permitting 'inter-operability' of the components of
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 183

distributed systems that are constructed and operated in a decentralized

way. M o r e o v e r , in the special circumstances that characterize markets in
which network externalities are an important consideration for vendors and
purchasers, competition between variant product designs in emerging areas
of technology has directed the attention of industrial organization econom-
ists and others to the private and social consequences of business strategies
that p r o m o t e or inhibit de facto standardization through the formation of
' b a n d w a g o n s ' (see, for early examples, Farrell and Saloner, 1985, 1986a,b;
and David, 1985, 1987). But while these issues have been very much in the
limelight of discussion in the industry press and the professional economics
journals recently, the standard-setting activities focused on questions of
guaranteeing m i n i m u m quality, safety, and environmental protection have
not languished in darkness. Indeed, they continue to expand in volume and
in the range of industries on which they impinge (see, for example, Hawkins
et al., 1994). 1 It seems appropriate, therefore, to return some analytical
attention to the broader, m o r e general questions raised by the role of
standardization in the d e v e l o p m e n t of new technologies and the perform-
ance of the industries that are based on them. At least, that is our intention
in this paper.
T h e analysis presented here re-addresses what may be taken to be the
classic, generic p r o b l e m of determining how much uniformity (or, equiva-
lently, how much diversity) it is best to maintain during the d e v e l o p m e n t of
a new industrial technology. Insofar as our concern is with when and how to
set standards, and inasmuch as standards may be construed as information
products, there necessarily will be some attention given to the externalities
that arise from the production and distribution of new technical knowledge.
But ' n e t w o r k externalities' and their implications do not emerge as focal
points of the discussion here. F u r t h e r m o r e , as our analysis will envisage the
decision regarding the degree of standardization being m a d e and im-
p l e m e n t e d centrally, by an agent or agency responsible for an ensemble of
distinct production facilities, it can be said to most directly address the
p r o b l e m s of strict, de jure standardization, or the setting of technical
regulations----either by a public authority, or by the central m a n a g e m e n t of a

1The term 'standard' sometimes is used to refer to a document, which can be conceptualized
as an information product, and, in other contexts, to refer to the technical specifications or
operating characterisitics of tangible, physical commodities of varying degrees of complexity.
Generically, a 'standard' is to be understood, for the present purposes, as a set of technical
specifications that may be adhered to by a producer, either tacitly or as a result of a formal
agreement. It is helpful to distinguish among several kinds of standards----reference, minimum
quality, and interface or compatibility standards. See David (1987) for the introduction and
discussion of this taxonomy.
184 P.A. David, G.S. Rothwell / Int, J. Ind. Organ. 14 (1996) 181-201

p r i v a t e c o r p o r a t i o n . 2 A n d , as for t h e s e v e r a l p u r p o s e s o r s o u r c e s o f
e c o n o m i c b e n e f i t t h a t s t a n d a r d i z a t i o n is i n t e n d e d to a c h i e v e , we find it m o s t
r e l e v a n t f o r t h e p r e s e n t analysis to t a k e n o t i c e o f t h e w a y t h a t t h e s e affect
t h e t e m p o r a l d i s t r i b u t i o n o f t h e gains. Since w e a r e n o t c o n c e r n e d h e r e with
d e f a c t o s t a n d a r d i z a t i o n t h r o u g h m a r k e t c o m p e t i t i o n u n d e r c o n d i t i o n s in
w h i c h t h e r e a r e significant n e t w o r k e x t e r n a l i t i e s , it is o f less i m p o r t a n c e to
stress t h e d i f f e r e n c e b e t w e e n t h e benefits o b t a i n a b l e f r o m e c o n o m i e s o f
s c a l e , a n d t h o s e f r o m t h e c o o r d i n a t i o n of t h e activities o f m a n y c e n t r a l i z e d
decision-agents.
T h e p a p e r is o r g a n i z e d as follows. In S e c t i o n 2 we r e - e x a m i n e t h e g e n e r a l
p r o b l e m o f s t a n d a r d - s e t t i n g in m a r k e t s w h e r e n e t w o r k e x t e r n a l i t y c o n d i t i o n s
may or may not constitute a dominant consideration. As the economic
significance o f s o m e o f t h e c e n t r a l g e n e r i c issues o f s t a n d a r d i z a t i o n p o l i c y
a r e s t a r k l y i l l u s t r a t e d b y r e f e r e n c e to t h e c o n t r a s t b e t w e e n t h e w a y t h e
n u c l e a r p o w e r i n d u s t r y has b e e n d e v e l o p e d in t h e U n i t e d S t a t e s a n d in
F r a n c e , w e briefly r e v i e w t h a t e x p e r i e n c e in S e c t i o n 3. O u r p u r p o s e in d o i n g
so is to use t h a t c o n c r e t e case to m o t i v a t e t h e f o r m u l a t i o n of a s i m p l e a n d
r e a s o n a b l y g e n e r a l d y n a m i c p l a n n i n g m o d e l for t h r e e stages o f t e c h n o l o g i c a l
d e v e l o p m e n t a n d i n d u s t r i a l c a p a c i t y f o r m a t i o n . S e c t i o n 4 discusses l e a r n i n g
a n d s t a n d a r d i z a t i o n . T h e m o d e l , which we p r e s e n t in S e c t i o n 5, allows for
both incremental innovations based on localized learning from plant oper-
ating experience, and radical design changes involving the selection of
d e s i g n s for n e w p l a n t c o n s t r u c t i o n b a s e d on k n o w l e d g e g a i n e d t h r o u g h
comparison of the performance of alternative technologies. The situation of
t h e n u c l e a r p o w e r utilities, e s p e c i a l l y t h o s e in t h e U n i t e d S t a t e s w h e r e such
p l a n t s f o r m p a r t o f b a s e - l o a d c a p a c i t y , is c o n v e n i e n t for o u r p u r p o s e s in t h a t
it justifies c o n c e n t r a t i n g a t t e n t i o n o n s u p p l y - s i d e issues a n d i g n o r i n g c o m p l i -

2 Just as there are many different types of standards, standardization as a process can come
about in many ways, and therefore standards can appear in various forms. Following the
classification system proposed by David and Greenstein (1990), distinctions can be drawn
among the following: (a) unsponsored standards, these being sets of specifications that have no
identified originator holding a proprietary interest, nor any subsequent sponsoring agency, but
nevertheless exist in a well-documented form in the public domain; (b) sponsored standards,
where one or more sponsoring entities holding a direct or indirect proprietary interest--
suppliers or users and private cooperation ventures into which firms may enter--creates
inducements for other firms to adopt particular sets of technical specifications; (c) standards
agreements arrived at within, and published by, voluntary standards-writing orgalfizations; (d)
mandated standards, which are promulgated by government agencies that have some regulatory
authority. The first two of the foregoing emerge from market-mediated processes and are
referred to generally as de facto standards, whereas the latter pair usually issue from political
('committee') deliberations or administrative procedures that may be influenced by market
processes without reflecting them in any simple way. Both of them are often tagged loosely as
de jure standards, although when the standards actually do have the force of law, as in case
(d), it is convenient to refer to them as technical 'regulations'.
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 185

cations arising in other industries from considerations of demand. Further-

more, the nuclear power industry is one whose regulated status makes it
particularly relevant to examine the question of public policy affecting the
degree and timing of standardization achieved through technical regulations
imposed by a monolithic agency---even though, in the United States, in
addition to the regulatory authority vested in the federal Nuclear Regulatory
Commission (NRC), state utility commissions can establish policies and
Docedures that indirectly impinge on technological choices. In Section 6 our
model is analyzed for the purpose of answering the question: 'What is the
optimal, present value cost-minimizing degree of diversity in design among
the plants to be built in the industry's first wave of capacity formation?'
Section 7 concludes with a discussion of the implications of our results,
points out some of the limitations of the simplified analytical structure we
have examined, and suggests further lines of work that would approach
these questions, as we have attempted to do here, in a framework that
recognizes the essentially dynamic nature of the issues that are at stake.

2. Tensions between uniformity and diversity as sources of industrial

progress

The kernel of the problem posed for private and public decision-making
with regard to the setting of technology standards can be construed to be
nothing more, and nothing less than the fundamental issue with which all
social organizations are confronted: where to position themselves on the
terrain between the poles of 'order' and 'freedom' (see David, 1994, on
which the following draws heavily). From 'order' one can derive greater
predictability, the perfection of performance through repetition and routini-
zation, the economies of simplification, and other, kindred advantages.
Standardization as an intentional act or as an unintended, de facto outcome
of the interplay of actions taken for other purposes, creates order by
reducing variety. Order offers reduced uncertainty and thus permits
economizing on the costly gathering of information. The establishment of
simplified routines is known to be conducive to the attainment and
enhancement of many specialized skills. From all these effects will flow
immediate gains in productive and allocative efficiency. Such is the conven-
tional wisdom.
But, there is another side to the matter, as many have been just as quick
to point out. For one thing, to impose order in the shape of uniformity is to
risk imposing error on a larger scale, and ultimately on a scale that could
deny the possibility of exposing that error. For another thing, diversity may
be intrinsically desirable. Just as people may have inherent preferences for
the placidity that accompanies order, so they may have demands for intrinsic
186 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

novelties as means for combatting the malaise of boredom. Some years ago,
Scitovsky (1976) sketched an arresting portrait of modern-day, affluent
consumers as engaged forever in mixing novelty with repetition in largely
vain efforts to maintain a physiologically comfortable equipoise between
states of excess stimulation and states of inadequate arousal. Individuals are
not the same in their tastes for variety, any more than they agree on the
specifics of the pursuits and material objects that give them most satisfac-
tion. So, it is proper to acknowledge that the benefits of 'order' that can
achieved by technical standardization and homogenization may bring ef-
ficiency gains only at the cost of forgoing some of the economic welfare that
consumers (and users of technologies) would gain from being able to choose
from among a wider variety of product and process offerings (see Braunstein
and White, 1985, for a discussion of variety in the computer hardware
market).
That, however, does not end the accounting of the costs that could
potentially be laid at the door of the 'order' that comes from standardiza-
tion. The tyranny of systematization and uniformity can be charged also
with stifling creativity, and with constricting the scope for learning and
progress via experimentation and the selection of superior variants. True,
one should qualify this charge by noticing that order may serve to focus
experimentation in useful directions, and that simplification will often assist
learning via the isolation and identification of cause-and-effect relationships.
Such effects can contribute directly to improving the productive efficiency of
established routines. By the same token, however, the imposition of order
must circumscribe what can be discovered: it delimits the sphere of
observation and the domain over which selection is able to operate. In
reducing diversity, standardization curtails the potential for the formation of
new combinations and the regeneration of variety from which further
selection will be possible. Diversity, then, is the fuel that propels evolution-
ary adaptation, and, as the biologist Richard Lewontin (1982, p. 151) has
nicely put it: "selection is like a fire that consumes its own fuel...unless
variation is renewed periodically, evolution would have come to a stop
almost at its inception". And this is no less true of dynamic processes in
which selection is made deliberately by rational, optimizing agents than it is
of the blind workings of Darwinian 'natural selection'. (See Weitzman, 1992,
for a recent treatment of the economic value of diversity in such contexts.)
Perceptions of the polar opposition between these two conditions--
uniformity and freedom for variation from common norms--has led quite
naturally to the conceptualization of a force field existing between them,
within which it may be possible to define an optimal situation and alignment
for any particular form of public or private policy intervention, organization-
al structure, or institutional design. There has been a longstanding quest for
a golden mean, or ideal balance between freedom and order in political
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 187

affairs, and political economy more specifically. Following on from such

habits of thought, it is natural enough for economists to wish to frame the
question of standardization in regard to technologies and organizational
designs by asking how much adherence to uniformity would be 'best', either
in the sense of optimal for the economic welfare of the society as a whole, or
for various private interests within the whole.
The conventional juxtaposition of order against freedom in political
discourse historically has encouraged the simplistic view that the crux of the
problem lay in the fact that standardization and technological innovation
inherently are antithetical processes. That is the message that emerged, for
example, from the classic study of the rationalization movement in German
industry during the 1920s, by Brady (1933, p.26):
' Contrasted with the dynamic, revolutionary, stimulating power of science
and technological change, they [standards] represent stability, order, routine
and regularity. Standards lose their value with the necessity of change.
While they should be established in keeping with the dictates of scientific
development, still theirs is the empire of the relatively fixed and changeless
in a technically regimentized world. '
Brady was led to this striking summary formulation under the influence of
what might best be described as the 'philosophy' of the German Standards
Committee (Deutscher Normenausschuss) conveyed by the pronouncements
of Hellmich in 1927:
' Freedom and order are the two poles between which every controlled
development must run. Innate human creative and formative powers crave a
condition of freedom; reflective and systematizing reason demand order.
Both have their justification; only the unfettered interaction between these
two coercive forces will give salutary results. Standardization becomes
dangerous and harmful when it immoderately narrows the effective range of
creative fancy and thereby crushes the most powerful driving force to
progress. It dare, consequently, be extended only into realms where
development has practically run its course and where the methods and the
knowledge, providing a basis for rational ordering and control, lie readily at
hand .... (Hellmich, DIN, 1917-1927, Deutscher Normenausschuss, Berlin,
1927, transl. Brady, 1933, p.25). '
There is much with which one might agree in the foregoing statements,
but, some important demurrers and qualifications are in order. The first
caveat is that it is overly simplistic for modern policy-makers to accept this
juxaposition of 'freedom vs. order' in place of a careful examination of the
way that standardization affects the generation of innovations. Standardiza-
tion may well promote faster technological advance in some forms, par-
ticularly, as has been suggested, incremental improvements achieved by
pooling experience gained in production operations, that is, learning-by-
doing and learning-by-using.
188 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

Secondly, present-day industrial conditions relating to information net-

work technologies make it especially important to recognize that much of
the work of standards-development organizations is directed toward the
attainment of 'coordination' for purposes of innovation as well as immediate
productive efficiency. Formerly, the emphasis economists placed on the
putative existence of a trade-off between the dynamic efficiencies (from
creative freedom), and the static efficiencies (gained through uniformatiza-
tion and routinization) was certainly germane in assessing the work of
standards-writing organizations like the German DIN, or the economic
welfare consequences of the 'industrial rationalization' movements carried
on by cartels and trade associations in the Interwar Era. And it still may
have a legitimate claim to a central place in studies gauging the impact of a
modern minimum quality and safety standards, and environmental regula-
tions (see, for example, Hemmenway, 1975; Hawkins et al., 1994). This way
of looking at things, however, would seem to be much less appropriately
applied to the class of 'compatibility' or 'interoperability' standards that are
crucially important for the development of network technologies. Where
technological innovation is decentralized but there are strong complemen-
tarities and elements of technical interrelatedness in the design and en-
gineering of complex distributed systems, the conventional wisdom about
standardization impeding progress is likely to be dangerously misleading.
(See David, 1994, for further development of this argument.)
Yet a third, and a still more general note of qualification is the one most
immediately relevant for the analysis developed here: the simple 'freedom
vs. order' metaphor distorts perceptions of reality by recasting a problem
that is essentially dynamic into a static choice framework. Rather than
construing the tension as one that involves choosing under conditions of
imperfect information between the course that risks making an erroneous
standardization choice, and the one that forgoes the possibility of gaining
the benefits of a correct choice of technology on as wide a scale as possible,
one should consider the option of waiting until additional relevant in-
formation accumulates; and beyond that, experimenting with alternatives in
an effort to learn more. But, instead of tolerating whatever diversity would
arise in the absence of placing restraints on technological freedom of choice
at the very outset, there is the possibility of seeking some middle ground,
some optimized dynamic path for the degree of diversity (or its complement,
the degree of standardization), as the expected value of accumulating
information (learning) from experience changes in relation to the expected
value of moving to capture the putative benefits of applying what has been
learned on the widest feasible scale.
Under what conditions would such optimal standardization paths exist?
And, if they do exist, what if anything can be said about their general
shapes? These are the questions we shall try to answer in the context of a
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 189

quite specific model. Before turning to that task, it is instructive to glance at

one industry in which standardization of production facilities has come to be
widely regarded as important for economic performance, and yet, which has
had a history of wide international differences in the durations over which
technological diversity has been allowed to persist, and in the degrees of
internal technical standardization that eventually have been imposed on
producers by governmental regulation. This is the case of the nuclear power
industry, in which, at least among those countries that have made extensive
investments in this technology for electricity generation, the experiences of
the United States and of France exhibit the polar extremes of policy
regarding standardization.

3. Standardization and the nuclear power industry in the United States

and France

In discussions of industrial progress it is sometimes asserted that the slow

approach to standardization in production methods and product designs, or
the maintenance of a high degree of technological diversity, played a
prominent role among causes of success---or of failure--in one or another
line of business. But there is no modern instance to match the strength of
the consensus of expert opinion tracing the poor commercial performance of
the U.S. nuclear power industry in its failure to standardize at the right
point in the development of its capital-embodied technology. Many of the
present difficulties facing this industry are attributed the extreme hetero-
geneity of the population of plants that presently are being operated by the
nation's many electric utilities, and, even more than that, the lack of
standardization within (let alone among) each of the successive 'tranches' of
nuclear power plants whose design and construction were undertaken at
more-or-less the same points in time.
In the optimism for commercial nuclear power fostered by the U.S.
Atomic Energy Commission's (AEC) Power Demonstration Reactor Pro-
gram during the 1950s and early 1960s, several reactor prototypes were
completed before 1963, when the first order for a nuclear reactor was placed
by a U.S. utility for a plant at Oyster Creek, New York, and the program of
federal subsidization for technology development ended. The light-water
reactor (LWR) technology furnished by General Electric was the one to
which the AEC had devoted the bulk of its research and development
resources, and it was selected on the ground that it was the cheapest of the
available alternatives at that time. In the 'bandwagon' of utility contract-
signing for LWRs that ensued in the United States during the latter half of
the 1960s (see Cowan, 1990), many different plant designs, based on either
the Pressurized Water Reactor and or the Boiling Water Reactor, were put
190 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

into construction. Concerns over the issue of standardization arose in the

industry from the seeming perversity of the 'customization' of these plants,
despite the unanimity of the commitment made to the nuclear steam supply
systems (NSSS) belonging to the LWR family. There were four major
vendors of NSSS's, 16 architect-engineers designing the balance of the plant,
20 general contractors (eventually) responsible for the construction work,
and 50 utilities ordering plant to their specifications. The resulting com-
binatorial possibilities, combined with changes in regulations, made it
almost inevitable that---except for units constructed concurrently on the
same site---each of the facilities would be in a purely technical sense a
unique entity. It was already clear as early as 1970 that the resulting
diversity had created enormous regulatory problems, as a different safety
analysis was required before each unit could be commissioned. In addition,
there were numerous singularities in the design flaws that lengthened the
construction delays and escalated the costs (see Thomas, 1989, pp. 33,
60-74).
Far from being reversed in the course of the 1970s, this situation was
perpetuated in the constructing of a new and larger wave of plants. Average
times to complete construction doubled, from under 60 months in 1968 to
more than 120 by 1980, and largely due to the entailed ballooning of
construction costs, the real costs per kilowatt of capacity tripled, from $400
to $1200 in constant 1980 prices. Operating performance was disappointingly
low, far below expectations: between 1976 and 1985 the industry was
managing to generate only 63% of potential kilowatt hour output, despite
the fact that the high fixed costs implied that the plants should be kept up
and running as close to full capacity as possible (see Rothwell, 1990).
The French have been notable in avoiding following the U.S. nuclear
industry's high diversity path, and have been spared what appear to be its
painful consequences--even though they wholeheartedly embraced Ameri-
can LRW technology when they embarked on the large-scale commercial
exploitation of nuclear energy as a source of electricity. The Government of
France faced the problem of choosing between several designs available in
the late 1960s and early 1970s. Impelled by the importance which the
DeGaulle regime attached to avoiding dependence on the United States for
enriched uranium fuel, they had pressed ahead with the development of
their own design using natural uranium and gas-graphite reactor technology.
The coincidence in 1969 of strong economic considerations favoring rapid
expansion of the commercial nuclear power capacity of Electricit6 de France
(EdF) with the succession of Pompidou as the head of State, brought this
phase of experimentation to an abrupt end. To take advantage of the far
more extensive international commercial experience in the construction and
operation of LWRs, the French launched a program of building LWRs
under a license negotiated with Westinghouse. France's commercial nuclear
power program was characteristically centralized. There is only one electric
 P.A. David, G.S. RothweU / Int. J. Ind. Organ. 14 (1996) 181-201 191

utility, Electricit6 de France--EdF. There is but a single reactor supplier,

Framatome, and a single turbine-generator supplier, Alsthom. There has
been only one national government commission regulating the atomic
energy industry, the CEA, throughout the entire history of the use of
nuclear technology in France; local regulatory controls over nuclear power
plants are very circumscribed.
This monolithic structure has allowed France easily to implement a policy
of standardization not only in reactor designs but in the specification of the
steam turbines and the rest of the equipment and structures required by
electric power plants. A set of measures of the degree of standardization in
the complex, multi-dimensional facilities (see David and Rothwell, 1991,
1995) suggests the magnitude of the contrast with American experience: by
the time both countries had commissioned 40 plants, the index of stan-
dardization for France was 80% higher than that for the United States. And,
when both had arrived at the 60 plant mark, the French index had edged up
farther, to stand at 93% higher than the stationary U.S. level of plant
standardization. Similar contrasts between France and the United States in
the dimension of industry performance are widely remarked upon. It is
frequently observed that the French build nuclear power stations quickly
and cheaply, operate those plants in a technically demanding regime of
'load-following' without major safety problems, and introduce successive
technical improvements in a uniform manner (see, for example, Thomas,
1989). Nevertheless, one is entitled to wonder whether it is so obvious that
the American indulgence in diversity is necessarily mistaken, and that the
French path of rigorous standardization from the outset is always the
economically soundest course to follow in the development of industrial
capacity based on a new technology. Each course, it will be seen, may have
something to offer toward reducing production costs in the long run.

4. Learning and standardization

The opportunity to derive economic benefits from learning through

standardization, whether from one's own experience or from the experience
of others, argues for greater uniformity in production facilities. Stan-
dardization, like other forms of 'experimental control', makes it easier to
identify empirical irregularities that point to underlying structural conditions
deserving further investigation. Therefore, it promotes the learning process
directly and widens the sphere of application for what has been learned.
Yet, there is one respect in which standardization is learning's enemy. The
point is obvious when one considers evolutionary adaptive processes as a
mode of learning. Selection, whether natural, or in designed experiments,
can only work when there is diversity. After a period of experimentation,
during which information can be collected, it is desirable to choose the most
192 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

efficient design (assuming the conditions during the experimental phase are
expected to persist) and standardize practice on that one.
Indeed, there is a large body of theoretical work on the problem of how
to play 'multiple-arm bandits' that shows the optimality of the general form
of strategy just described: gather information sequentially about the per-
formance of the various (stochastic) processes, but eventually pick one of
them (see Cowan, 1987, 1991). For two- or n-armed slot machines, the
standard analysis treats each 'arm' as characterized by a stationary (albeit
unknown) probability distribution over the payoffs. Experimentation is
therefore non-intrusive in the sense that trying the alternative arms will in
no way alter the underlying odds of hitting the jackpot on any of them. But
non-intrusive experimentation is not a condition that is likely to be satisfied
when operating variant designs for durable industrial facilities.
The experimentation stage does not encourage controlled experiments:
plants that can be successfully operated under field conditions are likely to
be those about which more is learned. This positive feedback makes it more
likely that these plant designs will be favored as a basis for future
modifications, optimization of maintenance procedures, and operator train-
ing programs. In this fashion, design types favored by early circumstances
that were possibly adventitious could emerge as the ones on which in-
dividual practice should become standardized. Cowan (1990) has argued that
such a process led to the emergence of LWRs as a dominant technology in
the international nuclear power industry, although, it must be said, there
were political as well as economic and engineering efficiency considerations
that added to the complications of the story.
Therefore, although standardization might ease the improvement of
technical performance in an industry through learning processes, incremen-
tal learning creates positive feedbacks that can lead to de facto stan-
dardization. Also, the designs selected by decentralized sequential choice
need not be globally optimal, either because of present operating per-
formance or their susceptibility to future improvement. Just as the older
literature warned of the risk of governmental regulation leading to 'the
standardization of mistakes', so we must avoid attributing superior norma-
tive properties of efficiency to the de facto standardization that may emerge
from market competition among technologies under conditions of positive
feedback.

S. An intertemporal model of the standardization-diversity trade-off

To specify the ideas of the previous section, we consider a highly stylized

model of the way standardization decisions could affect the structure of cost
during the evolution of a complex technical system operated by a central
authority. We will simplify the problem by positing a three-stage evolution.
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 193

In Stage I, a set of plants is built; the decision is whether to standardize on

one or a few designs, or experiment by building many designs. In Stage II,
these plants are operated and undergo incremental improvements at rates
that are positively influenced by the degree of standardization achieved in
Stage I, so their operating costs are reduced. In Stage III, the original
complement of plants is replaced with an equal number of new ones. These
can be built to take advantage of what was learned during Stages I and II.
We will suppose, following the discussion above, that standardization aids
some forms of learning, while diversity promotes others. (Although we
propose a value of diversity function, we do not follow Weitzman, 1992.) Is
there an optimal degree of standardization for a central manager to adopt at
the outset that would balance the opposing effects on the present value of
expected production costs in the industry? We can find the answer to this by
formally representing the expected costs of each Stage as they appear ex
ante from the perspective of Stage I.

5.1. Stage I: Construction

The central manager's goal is to build N plants of types j =

1 . . . . , m, ..., M. For simplicity we assume symmetry: there are n plants of
each type, so nj = n. Therefore, we can write N = n . m . Of course, this
presupposes M i> N, that is there are enough variants to build the entire
complement on a one-of-a-kind basis. (The symmetry assumption is a
restrictive, special case; we address this in our conclusions.)
Next, we assume that construction costs are the same for the first plant of
each type: k0j = k 0. There is learning in construction activity. Because we
assume all plants are built instantaneously at the end of Stage I, where the
duration of Stage I is z~, learning is equivalent to economies of scale in the
n u m b e r of plants of a given type:
ki(n ) = k o n - r , 0 < T ~< 1, (1)
where 3' is a measure of learning. The total cost of each group of plants is
n . kj = n ( k o • n 1-*) = k o • n 1-~. The total cost of building all N plants is

K ( N I m ) = m ( k o n - ~ ) = m k O(kNm~/ ' - ~ = Nk0 ( -m~ ) . (2)

T h e r e f o r e , total cost is an increasing function of re~N, a measure of
relative diversity.

5.2. Stage H: Operations and incremental learning

Stage II costs are composed of two parts: a fixed charge and a variable
charge. Because Stage II begins with all plants complete and all plants are
depreciated by the end of Stage II, we assign the construction costs to the
194 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

beginning of Stage II. The fixed charge per plant is ko(m/N) ~. The variable
charge reflects the initial variable cost (with full capacity utilization) minus
the realized cost savings. We abstract from the dynamics of the learning
process, assuming instead that cost savings are realized instantaneously at
the beginning of Stage II. This abstraction depends on the assumption of full
capacity utilization. To simplify, we assume that the initial variable costs,
C0j, are equal to C O for all j. All plants start with the same initial cost. All
differences arise only from differential learning-by-using.
Because of incremental adaptations in design, retrofitting, and operating
procedures, operation results in learning opportunities that generate a
distribution of attainable cost savings. We are interested in only the largest
of these cost savings. If cost savings, ~ for plants of type j, are distributed
with mean/xj and variance o.2 the expected maximum cost improvement for
a single plant of type j is the mean of the cost-savings distribution:
E(Caxlnj = 1).
Now suppose by replicating the number of plants of type j that the size of
the sample drawn from the distribution, Vj(/xj, o-2) is increased to n j > 1.
How will E(u~ ax) be affected? Here, we make use of a general result on
extreme value distributions: when such samples are drawn from unimodal
distributions, the expected extreme value is a positive function of the mean
and the standard deviation. Also, it is a positive concave function of the
sample size. The effect of increasing sample size interacts positively with the
standard deviation of the underlying distribution. Allowing for one or more
plants of each type, the behavior of the expected maximum improvement
from experienced-based learning for type j can be described by

E(v?aXln):txi+B~(logn)', 0</3<1, forn~>l. (3)

(See Gumbel (1958)) and David (1974) for other applications of this
parameterization and its relationship to the extreme value distribution of
normal variates.)
In what follows we assume that o) = tr for all j and that/xj has a A(-'~, 0)
distribution. The mean across plant types of expected cost improvements for
any given plant type is ~ and its variance is 0. (We will relax the first of
these assumptions later.) The e x p e c t e d total cost improvement for each
group of n plants is
g(n) = ni.Li + Bern(log n)~ . (4)
The expected cost improvement for N plants for m types is

G(NIn ) = U[-~ + Bo.(log n)~]. (5)

Because we have assumed that nj = n and ~ = o., expected improvements
in the variable cost are the same. It is the distribution of /xj that is
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 195

responsible for differences in variable costs developing during Stage II.

Therefore, the variable cost per period is N [ C o - - ~ - Btr(log(N/m))~],
where Stage II has a duration of r 2 periods. These costs are discounted to
the beginning of Stage II by ~2(i, z2), where ~2 is a uniform series, present
value factor that depends on the interest rate, i, and z2. The expected cost in
Stage II (ECH) for the N plants is

ECII = N k o ( N ) V + ~2N[Co--~ - B~r(logN)~] . (6)

Greater diversity (re~N) drives up expected costs during Stage II. This is
from the unambiguously positive sign of the derivatives with respect to
(re~N) in Eq. (6).

5.3. Stage III: Second generation plants--construction and operation

In the third stage we suppose that the knowledge gained during Stages I
and II can be transferred to another generation of plant constructors, who in
Stage III will select one design on which to standardize. Stage III contains,
or represents, many later stages in which further design improvements might
be possible if complete standardization was not chosen in earlier stages.
There will be some point beyond which it will be optimal to suppress
diversity, as we pointed out in the discussion of the multiple-armed bandit
problems in Section 4. We simplify our model by identifying that point with
Stage III. We assume that the service life of first generation plants has been
suitably adjusted so they do not require replacement before the point at
which there is enough information to justify selecting a standard design.
Also we assume:
(1) The cost of building a single one of the N replacement plants built in
bin = k0" n-Y, where the
Stage III will be the average achieved in Stage I: .-j0
rate of learning is the same as in Stage I.
(2) Opportunities to observe different plants with different t%% make it
possible to select the best design, that is the maximal draw from the A(-~, 0)
distribution.
(3) Operation of the type of plant selected for standardization in Stage II
yields the same distribution of potential cost reductions, with the expected
maximum improvement exceeding that available in Stage II only if m > 1 in
Stages I and II. (If complete standardization was not achieved initially, there
will be additional gains from moving to complete standardization in Stage
III.)
Corresponding to assumption (1), the construction costs for N plants of a
single type built instantaneously at the beginning of Stage III will be

klII(N]/T/) ~- N [ k o ( N ) Y ] N -y , (7)
196 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

where N -~ represents learning in replacing all Stage I plants with a single

type. Corresponding to assumption (2), the selection of the type of plant
with the largest expected (single plant) cost reduction can be represented as
a draw from the extreme value distribution generated by A(~, 0). The
expected mean improvement is
E(/.t~IIIm)=~ + A O m ~, 0<a<l. (8)

This is where diversity during Stages I and II will have its pay-off by
permitting more scope for design learning, reducing operating costs during
Stage III. Variable cost per period is N [ C o - - ~ - AOrn"], where Stage III has
a duration of ~'3 periods. These are discounted to the beginning of Stage III
by the factor 63(i, 73). The expected cost in Stage III (ECIII) for the N plants
is
m
ECI, I = N k o ( - ~ ) N -r + 63N[C o - - ~ - AOm"] . (9)

6. Optimal diversity

What is the cost-minimizing value of m, the number of different plant

types to build in Stage I? To answer this question, we introduce discount
factors, 6,2(i, ~'1), which translates Stage II costs into Stage I present value
equivalents, and 6~(i, ~fl + ~'2), which translates Stage |II costs into Stage I
present value equivalents. Expressing all costs [see Eqs. (6) and (9)], we
write the present value expected costs at the beginning of Stage I,

c* lm/
+ 6 ;N1-2rko m~ + ~ 3~3N[C0 - ~ - A O m ~ ] . (10)
The first-order condition for a minimum (or maximum) is given by
b C * / a m = 0. The first derivative, OC*/Om, equals

qlTm ~'-1 + 6262q213 ( log N ) / 3 m l re,

- ~ 3o3q3otm
a-a ,
(11)

where
ql =- 6 2 N l - r k o + 8'3Nl-arko,
q2 -~ NB(r,
q3 =- NAO.
Note that q~, qa, and q3 are all positive. Next, we check the second-order
condition for a minimum: ~2C*/Om:==-6 > 0 . The second derivative is
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 197

, 2/" N\ -1]

+ 6;63q3m~-2(1 - a ) a . (12)

The first two terms on the right-hand side are negative, but the third term
is positive. Therefore, whether A is greater than, equal to, or less than zero
depends on where the second-order partial is evaluated. To evaluate the
expression for A where OC/Om = 0, we find from Eq. (11), multiplying both
sides by ( 1 - a ) m - l :

t~ 3 8 3 q 3 a ( 1 -- ct)m ~-2 = ql(1 _ a)ymV-2

+ 6262q2(1 -- a ) f l log m (lla)

Substituting this into Eq. (12) yields

N -1

(13)
T h e r e are two main cases:
Case I, where ( y - a)~<0, which implies that both terms in Eq. (13) are
negative and is thus sufficient to show that no interior cost-minimizing value
f o r m exists; and
Case II, where ( y - a ) > 0 is necessary, but not sufficient, for the
existence of an interior optimum for the value of m.
The interpretation of the results for Case I is straightforward and
conforms to our intuition. Because of the implications of increasing returns
(where the elasticity (y) of Stage III construction costs with respect to the
n u m b e r of plant types, m, built in Stage I is equal to or less than the
elasticity (a) of expected Stage III cost reductions with respect to m ) , C * ( m )
rises to an interior m a x i m u m . U n d e r these conditions there are two subcases
of interest: Case IA in which the lowest costs are achieved at one extreme,
where only one type of plant is built f r o m the outset (m = 1), and the degree
of relative standardization is 1 - (re~N) = [1 - (1/N)]---~ 1. In the other case,
Case IB, maximal diversity is superior from the viewpoint of cost minimiza-
tion and m* = N, so the degree of relative standardization is 1 - ( N / N ) = O.
Intuition suggests that when the benefits of learning how to operate plants
m o r e efficiently in Stage II are not very heavily discounted (when 6 ~ is not a
small number) and where the learning opportunities that depend on initial
diversity are more constricted (a is closer to zero than to one, but not less
than y, and 0 is small), Case IA would emerge. Standardization from the
outset would then be the optimal strategy.
198 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

These intuitions are readily confirmed. We can simply evaluate C*(N[m)

at the two extremes, m = 1 and m = N, and examine the difference:

C*(NIm = 1 ) - C*(NIm = N ) = 82koN(N -~ - 1 ) - 8 'zSzNBtr(log N)~

+ ~'3koNl-V(N -~ - 1)
+ ~ '363NAO(N ~ - 1). (14)

Since N > 1, the first three terms of this expression are negative.Therefore,
the only way for the expression to be positive (implying that the level of
costs is lower under maximal diversity), is for A, 0, a, and 8 383 to be large.
Next, we examine Case II in which (3"- a ) > 0 satisfies the necessary
condition for the present value of expected costs to be strictly minimized by
selecting some initial level of design diversity, i.e. is m > 1. For a strict
minimum, (3" - a) > 0 is not sufficient. The second-order condition must be
satisfied where OC*/Om=O or A > 0 . From Eq. (13), the second-order
condition is satisfied when
, N ~ -1
ql(3"-a)3"m'--~2~2q2fl(logm) [a+/3(log~-) ]. (15)

Solving the first-order conditions for m v from Eq. (11) and substituting
into Eq. (15), the condition for a strict minimum of C* is

3~''3q3a ~ (N)~( ( N ) -1] (16)

( 3 " - a ) 8 2 ~ 2 q 2 - ~ r n > log 3'+fl log-~- ,

with (3' - a) > 0.

Can a strict minimization of the present value of expected costs occur by
following an initial policy of maximal experimentation with m = N, i.e. with
no duplication of plant designs in Stage I? We have found that in Case I
such a policy will be better (than initial standardization) when A, 0, a, and
8 ; 8 3 a r e large. For Case II, however, the inequality in Eq. (16) can never be
satisfied at m = N, because with m = N the logarithm of N / m is equal to
zero and its inverse is undefined.
Not too surprisingly, a policy of complete standardization, where m = 1,
could be optimal for Case II. The second-order condition in this special case
is

,8 383 aAO
(3" - a ) ~ 2 ~ flBtr > (log N)~[3" +/3(log N ) - I ] , (16a)

which, given (3' - a), could be satisfied for moderate values of N if (aAO) >
([3Btr) and (8 383) > (8 '282). Also, the first-order condition at m = 1 can be
represented as
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 199

t P
6 3~30~ AO Yko N - v ~ 2 + ~ 3N - r
(logN)~ - ~82 /3 Bo"> /3B-~o'-- 6-~82 (17)

Therefore, if (A. 0) is large in relation to (B. tr), Eq. (17) could be satisfied
if N is large enough to make the negative term on the right-hand side
negligibly small. This implies that if the number of plants to be built is large
enough, a policy of initial standardization would be strictly optimal--under
Case II conditions--because the benefits gained in construction, and from
incremental improvements learned in operations, outweigh the foregone
gains from initial experimentation with alternative plant designs.

7. Conclusions and implications

Our analysis serves to emphasize the point that the issue surrounding the
question of technical standardization in the design of nuclear power plants is
fundamentally a matter of timing: eventually complete standardization will
be desirable. The problem is how soon one should move toward achieving
this. For example, without establishing the empirical facts about the strength
of learning effects in construction, operation, and redesign, we cannot state
that U.S. policy erred initially in encouraging and later in tolerating
diversity in the nuclear power industry.The accidents of history that brought
a halt to the construction of nuclear power stations in the United States
during the 1970s, and the persistence of political and economic conditions
that are likely to continue to block the construction of new nuclear capacity
until the next century (see Rothwell, 1993, 1994), have made more salient
the costs of extreme diversity and have deferred realization of whatever
benefits it might have yielded in the redesign process. In this respect the
United States is suffering from a lack of standardization.
Our analysis of Case I suggests that when the effects of learning through
diversity are strong, standardization can be an all-0r-nothing proposition
and the best course could turn out to be a total initial eschewal of
standardization. That particular conclusion is shaped by the special assump-
tions of our model, notably, by the symmetry assumption. The latter
assumption excludes the chance of getting the benefits of standardization in
construction and operational learning without having to sacrifice the oppor-
tunity to learn about the ability to improve variant plant types through some
degree of diversity. The results might be different were we to relax the
symmetry conditions, for example, by allowing n of one plant type and one
of each of the other plant types: N = n + ( m - 1). Even if learning for
redesign were considered, the extreme heterogeneity of the U.S. nuclear
power plant population represented a needless sacrifice of the benefits of
standardization. This illustrative restriction is only one of many that should
200 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201

be considered. T h e general combinatorial p r o b l e m of optimally distributing

N plants across m > N types is an analytically difficult o n e that lies b e y o n d
the scope of the present work.
W e have s h o w n that if we could p r o p e r l y gauge the degree of similarity or
diversity a m o n g complex, m u l t i - c o m p o n e n t p r o d u c t i o n facilities, we could
address the e c o n o m i c optimization issues associated with the d e g r e e of
s t a n d a r d i z a t i o n in an industry. While o u r m o d e l can be applied to the
n u c l e a r p o w e r industry, one could also apply the same c o n c e p t u a l frame-
w o r k to o t h e r large, technical systems. O n e could follow the same a p p r o a c h
in e x a m i n i n g the o p e r a t i n g characteristics of airline fleets, for the same
e c o n o m i c benefits to standardization are relevant for airplanes: m a n u f a c t u r -
ing costs, o p e r a t o r training, safety, and operating reliability. H o w e v e r , in
the airline example, one must consider the influence of c o m p e t i n g fleets. We
h a v e i g n o r e d this issue for nuclear p o w e r plants, i.e. the role of internation-
al t e c h n o l o g y transfer and c o m p e t i t i o n in determining the optimal n u m b e r
o f plant designs. B e c a u s e of this, and o t h e r simplifications, this p a p e r should
be c o n s i d e r e d as a first step in our u n d e r s t a n d i n g of the i m p o r t a n c e of
s t a n d a r d s in the dynamics of large technical systems.

References

Brady, R.A., 1933, The rationalization movement in German industry: A study in the
evolution of economic planning (University of California Press, Berkeley, CA).
Braunstein, Y.M. and L.J. White, 1985, Setting technical compatibility standards: An economic
analysis, The Antitrust Bulletin 30, 337-355.
Cowan, R., 1987, Backing the wrong horse: Sequential technology choice under increasing
returns, unpublished Ph.D. dissertation, Stanford University.
Cowan, R., 1990, Nuclear power reactors: A study in technological lock-in, The Journal of
Economic History 50, no. 3, 541-567.
Cowan, R., 1991, Tortoises and hares: Sequential technological choice under increasing
returns, Economic Journal 102, (July): 801-814.
David, P.A., 1974, Fortune, risk, and the microeconomics of migration, in: P.A. David and
M.W. Reder, eds., Nations and households in economic growth (Academic Press, New
York).
David, P.A., 1985, Clio and the economics of QWERTY, American Economic Review, 75(3):
332-337.
David, P.A., 1987, Some new standards for the economics of standardization in the information
age. In: P. Dasgupta and P. Stoneman, eds., Economic Policy and Technology Performance.
Cambridge University Press, London, ch. 9.
David, P.A., 1994, Standardization policies for network technologies---the flux between
freedom and order revisited, in: R. Hawkins, R. Mansell and J. Skea, eds., The politics and
economics of standards in natural and technical environments (Edward Elgar, London).
David, P.A. and J.A. Bunn, 1988, The economics of gateway technologies and network
evolution: Lessons from electrical supply history, Information Economics and Policy 3,
165-202.
 P.A. David, G.S. Rothwell / Int. J. Ind. Organ. 14 (1996) 181-201 201

David, P.A. and S. Greenstein, 1990, The economics of compatibility standards: an intro-
duction to recent research, Economics of Innovation and New Technology l, no. l&2, 3-42.
David, P.A. and G.S. Ruthwell, 1991, Performance-based measures of nuclear reactor
standardization. Center for Economic Policy Research Publication, no. 247. Stanford
University, Stanford, CA.
David, P.A. and G.S. Rothwell, 1995, Measuring standardization: An application to the
American and French nuclear power industries, European Journal of Political Economy
(forthcoming).
Farrell, J. and G. Saloner, 1985, Standardization, compatibility and innovation, The Rand
Journal of Economics 16, 70-83.
Farrell, J. and G. Saloner, 1986a, Installed base and compatibility: Innovation, product
preannouncements and predation, American Economic Review 76, no. 5, 940-955.
Farrell, J. and G. Saloner, 1986b, Standardization and variety, Economic Letters 20, 71-74.
Foray, D. et al., 1991, Information technology standards: The economic dimension (OECD,
ICCP-D5TI 25, Paris).
Gumbel, E.J., 1958, Statistics of extremes (Columbia University Press, New York).
Hawkins, R., R. Mansell and J. Skea, eds., 1994, The Politics and Economics of Standards in
Natural and Technical Environments. Edward Elgar, London.
Hemmenway, D., 1975, Industrywide Voluntary Product Standards. Ballinger Pess, Cambridge,
MA.
Kindleberger, C.P., 1983, Standards as public, collective, and private goods, Kyklos 36 (Fasc.
3), 377-396.
Lewontin, R.C., 1982, Human diversity (Scientific American Books, New York).
OTA, 1992, Global standards: Building blocks for the future (U.S. Congress, Office of
Technology Assessment, TCT-512, Washington, DC).
Rothwell, G.S., 1990, Utilization and service: decomposing nuclear reactor capacity factors,
Resources and Energy, 12: 215-229.
Rothwell, G.S., 1993, New nuclear power plants require utility structure changes, in: T.
Anderson and C. Forsberg, eds., Nuclear power as a large-scale global energy option (Oak
Ridge National Laboratories).
Rothwell, G.S., 1994, U.S. nuclear power policy to the year 2000, in; G. Hinman and T.
Lowinger, eds, Nuclear power at the crossroads (International Research Center for Energy
and Economic Development).
Scitovsky, T., 1976, The joyless economy: An inquiry in human satisfaction and consumer
dissatisfaction (Oxford University Press, Oxford).
Thomas, S.D., 1989, The Realities of Nuclear Power: International economic and regulatory
experience. Cambridge University Press, London.
Weitzman, M.L., 1992, On diversity, Quarterly Journal of Economics 107, no. 2, 363-405.